Design Optimization of Real-Size Steel Frames Using Monitored Convergence Curve

Abstract

It is an undeniable fact that there are main challenges in the use of metaheuristics for optimal design of real-size steel frames in practice. In general, steel frame optimization problems usually require an inordinate amount of processing time where the main portion of computational effort is devoted to myriad structural response computations during the optimization iterations. Moreover, the inherent complexity of steel frame optimization problems may result in poor performance of even contemporary or advanced metaheuristics. Beside the challenging nature of such problems, significant difference in geometrical properties of two adjacent steel sections in a list of available profiles can also mislead the optimization algorithm and may result in trapping the algorithm in a poor local optimum. Consequently, akin to other challenging engineering optimization instances, significant fluctuations could be observed in the final results of steel frame optimization problems over multiple runs even using contemporary metaheuristics. Accordingly, the main focus of this study is to improve the solution quality as well as the stability of results in metaheuristic optimization of real-size steel frames using a recently developed framework so-called monitored convergence curve (MCC). Two enhanced variants of the well-known big bang-big crunch algorithm are adopted as typical contemporary metaheuristic algorithms to evaluate the usefulness of the MCC framework in steel frame optimization problems. The numerical experiments using challenging test examples of real-size steel frames confirm the efficiency of the MCC integrated metaheuristics versus their standard counterparts.